Marketing Data Analyses, Part 1

Photo by Campaign Creators on Unsplash

Over the past year I had a great chance working with FB/Instagram data. I wanted to share a few cool things that I have learned and show some ways that others can display and dig into their own marketing data.

Note: Obviously I cannot show real brand ad data, so all the data here is made up (which is shockingly time consuming to do!). Additionally, in the below I will use the terms audience and ad set interchangeably, since many of the accounts I have seen often set it up this way. Not all accounts have different audiences in different ad sets.

Also, please, find the link to my Tableau analysis here.

As a first step, I want briefly introduce some of the marketing business metrics that I used for my analysis:

• CPA (Cost per action) — in this case customer acquisition cost (CAC) or the amount of money it costs to acquire a new customer
• CPM (cost-per-thousand impressions) — is the bid paid per one thousand views (impressions) when running ads on FB/Instagram
• CPC (cost-per-click) — is the amount paid every time a user clicks an ad. Please note that many of these metrics could also be broken down in terms of “unique” clicks or unique views, but for simplicity I am just keeping it all as clicks for now.
• CTR (click-through-rate) — shows the percentage of people who see an ad and click through to the post-click landing page
• Conversion rate — in this case refers to the number of people who purchased divided by the number of people who clicked on an ad (e.g. if one in 10 people who click purchase, that implies a 10% conversion rate. 1–2% is industry standard in D2C).

This first blog post shows a few ways to break down and analyze marketing data by based on specific audiences and ads (images, gifs, or videos). The overall goal is to be able to look at this data and then assess the following:

1. How your marketing performs across your various audiences, and across your marketing assets (images, gifs, videos, etc). Do certain images or videos perform better in all audiences or only in certain ones? For instance, if you notice that your larger audiences respond well to video, you can focus on creating more.
2. Better understand what drives your marketing performance within each audience and across each image. Is CPA higher in one because their CPMs are too high? Or does the audience simply click less (lower CTR) or convert less than others per click (conversion rate)? Do certain images catch people’s eyes better (higher click through rates) but ultimately fail to convert (conversion rates)?
3. See whether your ad account is completely dependent on certain audiences or images (note, for simplicity this analysis does not go into audience overlap).

Views 1 & 2: Overall view of key drivers of CAC by Audience

I start with the below views because they are a great way to see the overall health of the ad account as well as the performance within each audience. In one quick glance you can see where your spend is going, what the CAC per audience is, and then get a better understanding of what drives each CAC.

For instance, in the below you can easily see that Audience 3 is underperforming in terms of CAC. Looking to the right you can get a clear picture of why — it starts off with a decent CPM, but then has a lower CTR, which leads to a higher CPC than on average. Next, those clicks are also not converting as well as other campaigns, leading to this high CAC.

Next, I have the exact same view, but this time each Audience also breaks out performance by ad image. This lets you get an even more in-depth view of what drives each return.

For example, this view lets you see that while Audience 6 is doing very well on the whole, its returns are largely driven by Image 4, and that Gif 3 is actually damaging that set’s overall return. Based on this you may want to remove Gif 3 and replace it with another asset, perhaps one that is similar to Image 4.

The only downside of the above is that they analyze metrics in a specific time period, and does not show how performance has changed over time, which I will show in my next blog post.

View 3: Distribution of spend and returns by audience and ad.

Another view I created that I like a lot shows the spend more visually, and compared to the previous views helps you highlight your dependence on an individual audience (and images within each). Each category is colored based on their CPA, with green showing overperformers and red showing underperforming areas. The below view lets you easily see that Audience 6 (ad set 6) receives the most spend, and that generally speaking it also has a better CPA than other ad sets. Audience 3 in the bottom right, however, generally has worse performance, and lower spend. You could also color these based on target CPAs to see how they perform against your target as opposed to each other (e.g. red could mean worse than target and green better than target).

This view also lets you see that Audience 8 actually has some strong images and potential for success, but is being weighed down by Image 2. Perhaps that image does not resonate with that audience, and you should consider removing it or replacing it with another one.

View 4 and 5: Changing the focus to examine ad performance

Similar to the first two views above, I ran the same analysis but this time by ad as opposed to audience. This lets you easily see each ads individual performance, regardless of audience. Compared to the above, this also lets you see how distributed your spend is by ad, with Image 4 and Video 6 taking a large amount of the spend in this example.

Next, similar to above, I then break out each ads performance across audiences. This lets you see which audiences each ad resonates with, and then even think about how to create new ads for specific audiences. For example, Video 6 does fairly well except with Audience 10, and this chart helps you see that this audience has a lower CTR and a lower conversion rate than others for this image. This also helps you think about why each one may or may not resonate with an audience. Is Audience 10 older, and thus not like videos? Or does the video show something that might not go well with them?

View 6: Purchase source by image

Finally, I examine transaction source by ad. This view is similar to the above, but because it shows transactions by size, it helps you visually understand your dependency on certain ads. If your ad account is completely dependent on one or two images, it may be time to start testing more so that you are prepared if any audience gets oversaturated with an image. It also helps you visualize whether a single audience dominates one ad’s spend.

Conclusion

In this post, I walk through a few helpful ways of visualizing your marketing data by audience and image. I wanted to share these views because I think they’re great at not just showing you what the performance is, but also explaining what drives that performance. Additionally, many accounts are overly reliant on individual audiences or ad assets, and some of these views can help brands decide to try and test more images, and furthermore use existing data to help them understand what type of images might work best in certain audiences.

My next post will be on analyzing marketing data changes over time.

Let me know if you have any comments or know of any other cool ways of visualizing marketing data!